Poverty and conflict in Thailand’s Deep South

Thailand’s so -called Deep South has experienced much deadly violence since the early 2000s. This article investigates its determining factors in the context of the larger civil unrest/civil war literature — work on Southeast Asia being sparse and work on Thailand almost non-existent. The focus is on 37 sub-provincial districts of four of Thailand’s 77 provinces covering the years from 2012 to 2019. Centering on descriptive statistics with additional panel regressions, it is found that reduced poverty incidence, increased educational attainment for males, and increased district-level per capita income are all associated with reduced conflict intensity (a smaller number of conflict-related deaths). In contrast, ethno-religious backgrounds and certain geographic features are not associated with either increases or decreases in conflict-related deaths. hailand’s

hailand's "Deep South" is predominantly populated by Malay-Muslim people (the majority of Thais identifying as Buddhists).Consisting of 4 provinces (out of 77), the region has been restive for many decades, and ongoing violence since the early 2000s has cost over 7,000 lives.This has resulted in a state of overt military presence, surveillance, and interference in daily life not seen anywhere else in the country.This article examines the region's politically motivated violence by drawing on the literature on the economics of civil war.It finds that reduced poverty incidence, increased educational attainment for males, and increased district-level per capita income are all associated with reduced conflict intensity as measured by a smaller number of conflict-related deaths.In contrast, ethno-religious background and certain geographic features are not associated with either increases or decreases in conflict-related deaths.
The next section of this article provides a political and economic background to the Deep South.This is followed by a brief review of the literature and descriptively and analytically tests its relevance to the Thai context.The final section concludes by reprising the key themes of this article.

Thailand's Deep South
Thailand's southernmost region spans the provinces of Pattani, Yala, and Narathiwat as well as some parts of Songkhla (see Figure 1).Often referred to as the country's Deep South, it is a region in which a tense relationship between Thai security forces and Malay Muslim communities can be traced back to the beginning of the 20 th century when the Muslim sultanate of Patani was forcibly incorporated into Thailand.With its distinctive historical background and ethno-linguistic and ethno-religious identity, the region is home to more than 2 million people, mostly Muslim, and accounts for about 3% of Thailand's Buddhist-centric total population. 1 Despite long-standing misgivings, it was not until early 2004, when a group of people seized a large number of weapons from one of Narathiwat province's military camps, that the simmering conflict escalated into open violence.According to the NGO Deep South Watch (DSW), between January 2004 and August 2020 around 20,000 incidents have been recorded, leaving about 7,000 people dead and 13,000 injured.The number of such incidents at first declined only gradually but then saw a sharp drop in 2008 due to a Thai policy of T increased military and special law enforcement presence (as well as increased security expenditure).An additional large decrease in incidents is observed after 2013 when a peace process was established (see Figure 2).
Of all the conflict-related incidents, those caused for identifiably separatist motives dropped from about 70% in 2013 to about 47% in 2019.Even though the total number of incidents is dropping, the large percentage with unidentified causes creates uncertainty among the regions' people, and skepticism over just who the perpetrators are.
A note on language, the term "incident" means incidents with casualties as well as casualty-free events.Further, the press, and at times scholars, refer to an "insurgency" and to "insurgents."These are politicallyladen terms since, as noted, much of the region was taken by force and "insurgents" are often simply Deep South local people whose ancestors inhabited the land before it was taken.That said, there are suggestions that at least some perpetrators have permanently fled or temporarily slipped across the modern-day, densely forested, and generally ill-patrolled Thai-Malay borderfrom where they return to stage attacks.
Annual per capita income in the Deep South is only approximately THB 77,000 (USD 2,400)-in Narathiwat province it is only THB 62,000 (USD 2,000).In contrast, Thailand's average gross provincial product (GPP) per capita in 2018 was THB 237,000 (USD 7,500).Figure 3 shows the incidence of poverty across three of region's four provinces (Songkhla is not included as some districts of this province are relatively rich).Narathiwat and Pattani have been among Thailand's 10 poorest provinces (out of 77) for more than a decade.Other provinces in the South are relatively prosperous, which has overshadowed the Deep South's need for economic assistance.That said, Figure 4 shows that the depth of the poverty in the Deep South has resulted in the South (as a whole) having the Kingdom's highest rate of poverty incidence (making little to no progress over the past 10 years).

Literature review
In 1998 Collier and Hoeffler published "On Economic Causes of Civil War" which became the touchstone for the ensuing "greed versus grievance" debate among scholars and policymakers.The article posits that low opportunity costs may prompt impoverished young men to join a rebel movement.This, in part, led to the formulation of a "conflict trap" theory that holds that poverty and conflict not only are inextricably linked but that they create a mutual feedback loop, resulting in: Destruction of infrastructure; elevated security risks; and declines in education, economic activity and well-being.The theory was further developed to include an increased likelihood of a given conflict's continuation, recurrence, escalation, and diffusion.The trap is most serious for low-income countries.
Despite the bi-directional relationship between poverty and conflict noted in the literature, this article is interested only in the direction from relative socioeconomic deprivation to violent conflict.The literature has pinpointed several relevant factors that are briefly discussed in the following subsections, namely poverty, general economic conditions, education, ethnoreligious-linguistic polarization, and geography.

Poverty
This article has indicated the high incidence of poverty in the Deep South.Among the pioneering studies pointing to poverty as one of the defining causes of civil war, Fearon and Laitin (2003)  finds that, on average, the odds of civil war onset increase by 41% for each USD 1,000 reduction in per capita income.More specifically, a country at the 10th percentile of per capita income has an 18% chance of civil war, a median per capita income country has an 11% chance, and only 1% for a country at the 90th percentile.Likewise, Do and Iyer (2010) finds that the number of casualties per 1,000 population during the Maoist conflict in Nepal was strongly associated with the poverty rate and related socioeconomic development indicators such as the infant mortality rate, literacy, and road length.Poverty, in particular, was a strikingly relevant indicator for both the emergence of conflict and death rates in Nepali districts.A district with a 60% poverty rate was 1.5 times more prone to reach 100  2014), however, concludes that while external income shocks can have an insignificant effect on the likelihood of instigating conflict, they affect the occurrence of conflict in the most trade-open regions.This point might be relevant to a tourist-dependent Thailand, but it does not apply to the Deep South, which sees little tourism (despite its natural beauty)-a point briefly addressed in the conclusion section. 5nstead, this article focuses on a broad GDP-related measure.Although there is a consensus that GDP per capita and conflict are inversely correlated, the mechanism that drives the correlation is still debated.Fearon and Laitin (2003) states that poorer countries do not have the resource to suppress conflict when it arises.Collier, Hoeffler, and Soderbom (2001) indicates that conflict is enabled by lower opportunity costs for individuals in states with lower per capita income.Between these positions, Dube and Vargas (2013), for example, finds that both mechanisms were present in the long-running Colombian conflict.It is plausible that the Deep South's long-standing and pervasive poverty as well as generalized low per capita income are related to the region's ongoing conflict.

Education
Good quality education equips people with knowledge and skills that widen their market opportunities.In general, educated people are more able than others to form their own views and judgements, thereby possibly reducing their scope for manipulation by conflict initiators.Agbor (2011) points out that individuals and societies can, in time, become more conflict resilient and that children can learn tolerance, respect differences, and become more empathic toward others.Similar to the Do and Iyer (2010) findings on literacy, Collier, Hoeffler,  and Soderbom (2001) finds that education measures and performance (such as an increase in male secondary school completion rates) are revealed to have significant conflict prevention or mitigation effects in the early stages of conflicts-i.e., an increase in education substantially reduces the expected duration of a conflict.However, the mitigating influence of education diminishes as conflicts wear on, possibly due to the educated demographic emigrating to non-conflict areas. 6ater this article illustrates the lack of male education in the Deep South and examines its relationship with conflict.Given that the Deep South is dominated by a Malay-Muslim population, whereas most of the remainder of the country is Thai-Buddhist, such polarization is included in this article's analysis.

Geography
The literature has come to routinely take geography and, especially, topography into account, the reason being that difficult-to-access territory can effectively serve as a safe haven for rebels to hide in, bide their time, and prepare future attacks.Researchers often include variables related to forest coverage and mountainous terrain.For example, Collier, Hoeffler, and Soderbom  (2001) states that a country with average forest cover experiences longer-lasting conflict than those with little forest cover (or mountains).Similarly, Fearon and Laitin  (2003) finds that a country that is about 50% mountainous (in the 90 th percentile), with median-level parameters otherwise, has a 13.2% chance of civil war onset over the course of a decade.An otherwise "identical" country has only a 6.5% chance on war onset if it is not mountainous (in the 10 th percentile).On a cautionary note, however, Pickering (2011) warns against using a broad statistical brush to paint all "mountain people" as bellicose. 7ithout sufficiently granular district level topographic/forestry data, this article uses the nature of a district's border as a proxy for provision of hidden insurgent access.

Data and methods
A panel dataset was constructed that included the 12 districts of Pattani, the 8 districts of Yala, the 13 districts of Narathiwat, and 4 of the 16 districts of Songkhla.The data covers the 8 years from 2012 to 2019.Data on conflict-related incidents are drawn from DSW.Further data came from the Information System for Basic Minimum Needs (BMN), compiled by the Department of Community Development in the Ministry of the Interior.Descriptive statistics were generated to provide summative background information and to provide some initial insights.Exploratory fixed-effects, randomeffects, and between-group means regressions helped to distinguish any unobserved characteristics that may affect conflict incidents in the Deep South (varying across districts as well as across time periods).The term "exploratory" is used due to the short timespan covered by the data.The resulting findings do not run counter to intuition. 8he estimated model takes the following form: (1) where Conflict intensity is the number of deaths per 1,000 people per district i (i = 1, …, 37) at time t (t = 2012, …, 2019).Xs reflects a set of characteristics, namely: Poverty incidence; education level of the male population; per capita income; an ethno-religious polarization index; and whether a given Deep South district is bordered by the sea and/or is adjacent to Malaysia.β0 is a constant term, β1 is a regression coefficient, u denotes a fixed effect of district i, and e denotes an error term that contains the effect on district i and time t.Note that except for time-invariant variables like geography, one-year lagged explanatory variables are used in accordance with the extant literature.

Descriptive statistics
Yala province has the largest number of incidents.Figure 5, however, shows that the number of deaths and injuries are highest in Pattani province, suggesting a more intense conflict situation there.In all, the 37 districts located in the Deep South experienced an average of 0.2 conflictrelated deaths per 1,000 people between 2012 and 2019, equating to about 9 deaths per year (see Table 1).Nevertheless, the variation in deaths across these districts is large, with the highest intensity of 1.28 deaths per 1,000 population, or 22 deaths annually, observed in the Kapho district in Pattani province in 2012.In contrast, the Nathawi district in Songkhla province has a near zero per 1,000 people (equivalent to 2 deaths annually). 9overty is measured as the percentage of households with annual household-based per capita income below THB 30,000 (USD 950) for 2012-2016 and THB 38,000 (USD 1,200) for 2017-2019.As already shown, the South as a whole is the poorest region of Thailand and, within the South, Pattani province is the poorest of all.The poverty surge observed in these provinces in 2017 is the result of the 2017 poverty income threshold increase.A steady reduction in the average poverty rates thereafter indicates some absolute improvement since then, but still not relative to the nation as a whole. 10Turning to education in the Deep South, less than a quarter of males have completed secondary education (or higher).The highest proportion is in Songkhla province (25.2%) and the lowest in Narathiwat province (22.1%).However, province-level averages hide substantial In sum, poor and un/undereducated males generate a potential pool of "conflict workers" due to the, presumably, low opportunity cost of engaging in conflict activities.This impression is reinforced by the income averages that reveal the Deep South is, again, Thailand's laggard.
In 2019, Narathiwat, Pattani, and Yala, respectively, had average per capita incomes of THB 56,436 (USD 1,792), THB 61,058 (USD 1,938), and THB 65,584 (USD 2,082), and thus were the first, the third, and the sixth poorest provinces-in contrast, the highest provincial per capita income is THB 152,801, while the average is THB 81,280. 11 As mentioned, the Deep South's majority population Notes: * See note 10, where "household per capita income" is explained.**The index is computed as 4 ∑si 2 (1si), where si is the proportion of religious group i in the population.
is Malay-Muslim.In contrast, their nation-wide population share is rather small (about 2 million, out of about 66.5 million people in Thailand).Surprisingly, however, the descriptive data show considerable variation in the ethno-religious polarization index, which measures the degree to which individuals in a population are "mixed" across different ethno-religious groups.Across the 37 districts, a minimum value of just 0.007 (almost no mixture) is found with a maximum value of 0.99 (a very thorough mixture).In particular, the four districts in Songkhla province in the dataset score average polarization indices as high as 0.91, implying far more diverse ethno-religious groupings there.Finally, geographic features like forest coverage and mountain ranges are mentioned in the literature as potential "hide-outs" and refuges for people perpetrating acts of violence.Unfortunately, while easily accessible satellite image maps show very densely forested mountains and sparse road networks across the Deep South, quantified data on forest cover or mountainous terrain at the district level are so limited that they could not be used.Instead, information on whether a district borders on the sea and/or shares a land border with Malaysia was used.There are plenty of media reports about insurgents slipping back and forth across the Thai-Malay border.Of the 37 districts, 10 are bordered by the sea and 12 are adjacent to Malaysia.12

Empirical results
Turning to the exploratory panel regressions, the number of deaths, injuries, and incidents against poverty incidence were plotted across time.A positive correlation between conflict and poverty is easily observed in the earlier years of the dataset, regardless of the choice of conflict variable, but the relationship vanishes toward the latter years (see Figure 6).This suggests the presence of time-variant factors underlying the poverty-conflict relationship.While the descriptive data show considerable variation across provinces and districts, data on time-variant district-level variables are limited.Nonetheless, exploratory panel regressions were attempted to learn, if nothing else, if they clearly refuted the descriptive impression about a link between relative socioeconomic deprivation, education, geography, and violent conflict.
A Hausman test suggests the use of a fixed-effects regression model as more appropriate than a randomeffects model.But since a fixed-effects model cannot deal with dummy variables, a between-regression model was also run to help capture the potential influence of geography in particular.So, in all, three models were run, a fixed-effects (within) regression, a random-effects (multiple) regression, and a between-group means (between) regression.Table 2 summarizes the results that, overall, suggest that a district with a low level of socioeconomic development in a previous year experiences intensified violence in a follow-on year (relative to districts that show a higher level of socioeconomic development and achievement). 13he between-group means regression in particular records a positive relationship between poverty incidence and the intensity of conflict.Specifically, a 1 percentage point increase in a district's poverty rate in a previous year is associated with an additional 0.032 deaths per 1,000 population in the following year.Given an average population of 50,203 persons per district, this equates to an increase of nearly 2 conflict-related deaths per year.Lagged per capita income and the current-year conflict-related deaths are negatively related.This quantitative finding is that a 1 percent increase in average per capita income level would lower conflict-related deaths per 1,000 population by 0.372, or 19 deaths per year.
The fixed-effects and random-effects estimations suggest a strong inverse association between educational attainment and the intensity of conflict (less education means more conflict).If the relationship is symmetric, this possibly suggests that by encouraging the male population to complete secondary education (or higher), a statistically significantly lower number of conflictrelated deaths might then be seen in subsequent years.However, depending on where one sets the cutoff point, this link is statistically insignificant for the betweengroup means model.This somewhat tenuous finding may be due to the rather small variation in male educational status across time.Also, information on specifically Islamic-based education, quite different from the dominant Thai-based education model, is not included in the dataset.
Including the ethno-religious polarization index into the models reveals that polarization in the Deep South is (probably) not a statistically significant variablecasting doubt on the oft-heard argument in national media that it is the ethnic background and different religious beliefs of those who live in the Deep South that are among the key factors responsible for the ongoing violence.The finding echoes those made in other country studies in the literature.
Finally, regarding geography, districts bordering the sea and/or Malaysia likewise does not affect conflict intensity.However, when including the geography variables in the random-effects between-group means regression models, the role of poverty incidence in determining conflict intensity in the following year becomes statistically prominent, unlike in the fixedeffects regression model where both geographical variables are omitted.This may suggest a biased, or at least inconsistent, poverty coefficient in this model.

Figure 6 :
Figure 6: Poverty incidence (in percent) and number of casualties, 2012-2019.Sources: Deep South Watch (DSW) and Information System for Basic Minimum Needs (BMN).
Number of casualties and incidents in the DeepSouth, 2004-2019.
Source: Deep South Watch (DSW).Pattani Yala Narathiwat South Figure 3: Poverty incidence by province (in percent), 2000-2018.Note: Detailed description of the data can be found in in Note 3. Source: Office of the National Economic and Social Development Council (NESDC, 2019).

The Economics of Peace and Security Journal
ConclusionTaking its main cues from the extant literature, this article explores factors governing the ongoing violence in Thailand's Deep South, the country's southernmost region.Using data from Deep South Watch and Thailand's Ministry of the Interior's Information System for Basic Minimum Needs, this study (exploratory as it is) is pioneering in that, to the best of the authors' knowledge, it is the first time such work has been carried out for the region-a region in which violence has cost many thousands of lives and continues to mar Thailand's tourism and direct foreign investment-dependent international image.While the number of violent events has declined over time, this apparent reduction in violence has been achieved by effectively garrisoning large population suppressing (if perhaps not repressing) the local population, and limiting education and investment opportunities.The Deep South is environmentally pristine and contains numerous archeological and contemporary Malay-Muslim cultural treasures.As such, it is a potential "hotbed" for both western-style ecotourism and for halal tourism (tourism ISSN 1749-852X https://www.EPSJournal.org.uk©EPS Publishing, 2020.All rights reserved For permissions, email: ManagingEditor@EPSJournal.org.uk